Reducing context memory requirements in a multi-tasking system

Abstract

A process for reducing the context memory requirements in a processing system is provided by a generic, lossless, compression algorithm applied to multiple tasks or multiple instances running on any type of processor. The process includes dividing data in a task of a multi-tasking system into blocks with each block containing the same number of words. For the data in each task, a word in a block having a maximum number of significant bits is determined, a packing width to the block of said maximum number of significant bits is assigned, and the least significant bits of each word in the block into a packed block of the packing width multiplied by a total number of words in the block is encoded with a lossless compression algorithm. A prefix header at the beginning of each packed block to represent a change in the packing width from the packed block from a packing width of a previous packed block is also provided.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS [0001] None FIELD OF THE INVENTION [0002] The present invention relates generally to reducing the context memory requirements in a multi-tasking system, and more specifically applying a generic, lossless, compression algorithm to multiple tasks running on any type of processor. BACKGROUND OF THE INVENTION [0003] Computer processors that execute multiple software functions (e.g., multi-tasking) using only on-chip memory must be able to operate the functions in a limited memory environment while conforming to size constraints of the chip and cost-effectiveness of manufacturing. While multitasking, a processor is simultaneously running numerous tasks that consume memory. Each task requires a certain amount of memory to hold each task's variables that are unique to itself. A problem with limited memory environments on processors is that all the memory is contained on the chip: the software operating on the chip does not use external memory. If mor...

AI Technical Summary

Benefits of technology

[0005] The problems of the prior art are overcome in the preferred embodiment by applying a generic, lossless compression algorithm to each task in a multitasking environment on a processor to reduce the context memory requirement of each task. The algorithm of the present invention operates as an adaptive packing operation. This method applies to any software system running on any type of processor and is useful for applications which process a large number of tasks and where each task consumes a significant amount of context memory.

Problems solved by technology

A problem with limited memory environments on processors is that all the memory is contained on the chip: the software operating on the chip does not use external memory.

If more memory is added, the chip requires a larger footprint and becomes more costly to manufacture.

For example, in a voice-data channel context, a barrier to increasing the number of channels per chip, and therefore reducing the power per channel and cost per channel, is the amount of on-chip memory that can be incorporated into a given die size.

However, this method does not compress variables or constants and uses software instructions instead of a faster system using a hardware engine.

Images